Background Integration of HIV services with other health services has been proposed as an important strategy to boost the sustainability of the global HIV response. We conducted a systematic and comprehensive synthesis of the existing scientific evidence on the impact of service integration on the HIV care cascade, health outcomes, and cost-effectiveness. Methods and findings We reviewed the global quantitative empirical evidence on integration published between 1 January 2010 and 10 September 2021. We included experimental and observational studies that featured both an integration intervention and a comparator in our review. Of the 7,118 unique peer-reviewed English-language studies that our search algorithm identified, 114 met all of our selection criteria for data extraction. Most of the studies (90) were conducted in sub-Saharan Africa, primarily in East Africa (55) and Southern Africa (24). The most common forms of integration were (i) HIV testing and counselling added to non-HIV services and (ii) non-HIV services added to antiretroviral therapy (ART). The most commonly integrated non-HIV services were maternal and child healthcare, tuberculosis testing and treatment, primary healthcare, family planning, and sexual and reproductive health services. Values for HIV care cascade outcomes tended to be better in integrated services: uptake of HIV testing and counselling (pooled risk ratio [RR] across 37 studies: 1.67 [95% CI 1.41–1.99], p < 0.001), ART initiation coverage (pooled RR across 19 studies: 1.42 [95% CI 1.16–1.75], p = 0.002), time until ART initiation (pooled RR across 5 studies: 0.45 [95% CI 0.20–1.00], p = 0.050), retention in HIV care (pooled RR across 19 studies: 1.68 [95% CI 1.05–2.69], p = 0.031), and viral suppression (pooled RR across 9 studies: 1.19 [95% CI 1.03–1.37], p = 0.025). Also, treatment success for non-HIV-related diseases and conditions and the uptake of non-HIV services were commonly higher in integrated services. We did not find any significant differences for the following outcomes in our meta-analyses: HIV testing yield, ART adherence, HIV-free survival among infants, and HIV and non-HIV mortality. We could not conduct meta-analyses for several outcomes (HIV infections averted, costs, and cost-effectiveness), because our systematic review did not identify sufficient poolable studies. Study limitations included possible publication bias of studies with significant or favourable findings and comparatively weak evidence from some world regions and on integration of services for key populations in the HIV response. Conclusions Integration of HIV services and other health services tends to improve health and health systems outcomes. Despite some scientific limitations, the global evidence shows that service integration can be a valuable strategy to boost the sustainability of the HIV response and contribute to the goal of ‘ending AIDS by 2030’, while simultaneously supporting progress towards universal health coverage.